Alexey Pechnikov wrote: > В сообщении от Wednesday 18 June 2008 23:40:05 John Stanton написал(а): >> Alexey Pechnikov wrote: >>> В сообщении от Wednesday 18 June 2008 18:42:25 John Stanton написал(а): >>>> The magic potion is the ability to embed Sqlite in the application >>>> server and avoid IPCs and multiple processes. >>> Why not multiple processes? And what about threads? If Sqlite library is >>> used in multi-threaded application server (for example, AOL Web Server) >>> we can use multiple read threads at same time, collisions may be with >>> write threads (I think, PRAGMA read_uncommitted=1 is usefull methode for >>> creating lock-free read queris, if it correspond to apptication logic). >>> And parallel reads performance is better then read and write in only one >>> thread. If "db timeout" command will work correctly (with Sqlite =<3.5.7 >>> this command not work right) then writes can wait for previos transaction >>> commit/rollback. And your opinion about this? >> Alex, >> This was the thinking. >> The overhead of spawning processes is avoided. Shared caching for >> Sqlite is assisted and POSIX locking overhead can be omitted. Open DB >> handles can be pooled. Synchronization can use fundamental primitives >> such as mutex. IPCs are avoided. Global storage can be used to advantage. > > Are you speaking about mutexes inside Sqlite or client programm mutex? What > is "Global storage"?
I use mutexes set up as read or write locks around Sqlite to synchronize access. Global storage is process memory within the scope of all threads and functions. > _______________________________________________ > sqlite-users mailing list > sqlite-users@sqlite.org > http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
